Several questions last Thursday on why the Kohonen plots may have a group in position (1, 1) in the training data and the same group in a different position in the testing data. There are two reasons this can happen:
1. the algorithm uses a random seed for the algorithm and with a different seed the order of the plot's nodes can change.
2. if the training and testing nodes have the same group showing a different degree of importance in the clustering algorithm the order can change.
From my reading of the modeler manual there is no easy way to change this.